summaryrefslogtreecommitdiff
path: root/util-linux/unshare.c
blob: d05cfdb6c62c742de6d668747b1c87253e69093b (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
/* vi: set sw=4 ts=4: */
/*
 * Mini unshare implementation for busybox.
 *
 * Copyright (C) 2016 by Bartosz Golaszewski <bartekgola@gmail.com>
 *
 * Licensed under GPLv2 or later, see file LICENSE in this source tree.
 */

//config:config UNSHARE
//config:	bool "unshare"
//config:	default y
//config:	depends on LONG_OPTS && !NOMMU
//config:	select PLATFORM_LINUX
//config:	help
//config:	  Run program with some namespaces unshared from parent.

// depends on LONG_OPTS: it is awkward to exclude code which handles --propagation
// and --setgroups based on LONG_OPTS, so instead applet requires LONG_OPTS.
// depends on !NOMMU: we need fork()

//applet:IF_UNSHARE(APPLET(unshare, BB_DIR_USR_BIN, BB_SUID_DROP))

//kbuild:lib-$(CONFIG_UNSHARE) += unshare.o

//usage:#define unshare_trivial_usage
//usage:       "[OPTIONS] [PROG [ARGS]]"
//usage:#define unshare_full_usage "\n"
//usage:     "\n	-m, --mount[=FILE]	Unshare mount namespace"
//usage:     "\n	-u, --uts[=FILE]	Unshare UTS namespace (hostname etc.)"
//usage:     "\n	-i, --ipc[=FILE]	Unshare System V IPC namespace"
//usage:     "\n	-n, --net[=FILE]	Unshare network namespace"
//usage:     "\n	-p, --pid[=FILE]	Unshare PID namespace"
//usage:     "\n	-U, --user[=FILE}	Unshare user namespace"
//usage:     "\n	-f, --fork		Fork before execing PROG"
//usage:     "\n	-r, --map-root-user	Map current user to root (implies -u)"
//usage:     "\n	--mount-proc[=DIR]	Mount /proc filesystem first (implies -m)"
//usage:     "\n	--propagation slave|shared|private|unchanged"
//usage:     "\n				Modify mount propagation in mount namespace"
//usage:     "\n	--setgroups allow|deny	Control the setgroups syscall in user namespaces"

#include <sched.h>
#ifndef CLONE_NEWUTS
# define CLONE_NEWUTS  0x04000000
#endif
#ifndef CLONE_NEWIPC
# define CLONE_NEWIPC  0x08000000
#endif
#ifndef CLONE_NEWUSER
# define CLONE_NEWUSER 0x10000000
#endif
#ifndef CLONE_NEWPID
# define CLONE_NEWPID  0x20000000
#endif
#ifndef CLONE_NEWNET
# define CLONE_NEWNET  0x40000000
#endif

#include <sys/mount.h>
#ifndef MS_REC
# define MS_REC     (1 << 14)
#endif
#ifndef MS_PRIVATE
# define MS_PRIVATE (1 << 18)
#endif
#ifndef MS_SLAVE
# define MS_SLAVE   (1 << 19)
#endif
#ifndef MS_SHARED
# define MS_SHARED  (1 << 20)
#endif

#include "libbb.h"

static void mount_or_die(const char *source, const char *target,
                 const char *fstype, unsigned long mountflags)
{
	if (mount(source, target, fstype, mountflags, NULL)) {
		bb_perror_msg_and_die("can't mount %s on %s (flags:0x%lx)",
			source, target, mountflags);
		/* fstype is always either NULL or "proc".
		 * "proc" is only used to mount /proc.
		 * No need to clutter up error message with fstype,
		 * it is easily deductible.
		 */
	}
}

#define PATH_PROC_SETGROUPS	"/proc/self/setgroups"
#define PATH_PROC_UIDMAP	"/proc/self/uid_map"
#define PATH_PROC_GIDMAP	"/proc/self/gid_map"

struct namespace_descr {
	int flag;
	const char nsfile4[4];
};

struct namespace_ctx {
	char *path;
};

enum {
	OPT_mount	= 1 << 0,
	OPT_uts		= 1 << 1,
	OPT_ipc		= 1 << 2,
	OPT_network	= 1 << 3,
	OPT_pid		= 1 << 4,
	OPT_user	= 1 << 5, /* OPT_user, NS_USR_POS, and ns_list[] index must match! */
	OPT_fork	= 1 << 6,
	OPT_map_root	= 1 << 7,
	OPT_mount_proc	= 1 << 8,
	OPT_propagation	= 1 << 9,
	OPT_setgroups	= 1 << 10,
};
enum {
	NS_MNT_POS = 0,
	NS_UTS_POS,
	NS_IPC_POS,
	NS_NET_POS,
	NS_PID_POS,
	NS_USR_POS, /* OPT_user, NS_USR_POS, and ns_list[] index must match! */
	NS_COUNT,
};
static const struct namespace_descr ns_list[] = {
	{ CLONE_NEWNS,   "mnt"  },
	{ CLONE_NEWUTS,  "uts"  },
	{ CLONE_NEWIPC,  "ipc"  },
	{ CLONE_NEWNET,  "net"  },
	{ CLONE_NEWPID,  "pid"  },
	{ CLONE_NEWUSER, "user" }, /* OPT_user, NS_USR_POS, and ns_list[] index must match! */
};

/*
 * Upstream unshare doesn't support short options for --mount-proc,
 * --propagation, --setgroups.
 * Optional arguments (namespace mountpoints) exist only for long opts,
 * we are forced to use "fake" letters for them.
 * '+': stop at first non-option.
 */
static const char opt_str[] = "+muinpU""fr""\xfd::""\xfe:""\xff:";
static const char unshare_longopts[] ALIGN1 =
	"mount\0"		Optional_argument	"\xf0"
	"uts\0"			Optional_argument	"\xf1"
	"ipc\0"			Optional_argument	"\xf2"
	"network\0"		Optional_argument	"\xf3"
	"pid\0"			Optional_argument	"\xf4"
	"user\0"		Optional_argument	"\xf5"
	"fork\0"		No_argument		"f"
	"map-root-user\0"	No_argument		"r"
	"mount-proc\0"		Optional_argument	"\xfd"
	"propagation\0"		Required_argument	"\xfe"
	"setgroups\0"		Required_argument	"\xff"
;

/* Ugly-looking string reuse trick */
#define PRIVATE_STR   "private\0""unchanged\0""shared\0""slave\0"
#define PRIVATE_UNCHANGED_SHARED_SLAVE   PRIVATE_STR

static unsigned long parse_propagation(const char *prop_str)
{
	int i = index_in_strings(PRIVATE_UNCHANGED_SHARED_SLAVE, prop_str);
	if (i < 0)
		bb_error_msg_and_die("unrecognized: --%s=%s", "propagation", prop_str);
	if (i == 0)
		return MS_REC | MS_PRIVATE;
	if (i == 1)
		return 0;
	if (i == 2)
		return MS_REC | MS_SHARED;
	return MS_REC | MS_SLAVE;
}

static void mount_namespaces(pid_t pid, struct namespace_ctx *ns_ctx_list)
{
	const struct namespace_descr *ns;
	struct namespace_ctx *ns_ctx;
	int i;

	for (i = 0; i < NS_COUNT; i++) {
		char nsf[sizeof("/proc/%u/ns/AAAA") + sizeof(int)*3];

		ns = &ns_list[i];
		ns_ctx = &ns_ctx_list[i];
		if (!ns_ctx->path)
			continue;
		sprintf(nsf, "/proc/%u/ns/%.4s", (unsigned)pid, ns->nsfile4);
		mount_or_die(nsf, ns_ctx->path, NULL, MS_BIND);
	}
}

int unshare_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
int unshare_main(int argc UNUSED_PARAM, char **argv)
{
	int i;
	unsigned int opts;
	int unsflags;
	uintptr_t need_mount;
	const char *proc_mnt_target;
	const char *prop_str;
	const char *setgrp_str;
	unsigned long prop_flags;
	uid_t reuid = geteuid();
	gid_t regid = getegid();
	struct fd_pair fdp;
	pid_t child = child; /* for compiler */
	struct namespace_ctx ns_ctx_list[NS_COUNT];

	memset(ns_ctx_list, 0, sizeof(ns_ctx_list));
	proc_mnt_target = "/proc";
	prop_str = PRIVATE_STR;
	setgrp_str = NULL;

	opt_complementary =
		"\xf0""m" /* long opts (via their "fake chars") imply short opts */
		":\xf1""u"
		":\xf2""i"
		":\xf3""n"
		":\xf4""p"
		":\xf5""U"
		":ru"	   /* --map-root-user or -r implies -u */
		":\xfd""m" /* --mount-proc implies -m */
	;
	applet_long_options = unshare_longopts;
	opts = getopt32(argv, opt_str,
			&proc_mnt_target, &prop_str, &setgrp_str,
			&ns_ctx_list[NS_MNT_POS].path,
			&ns_ctx_list[NS_UTS_POS].path,
			&ns_ctx_list[NS_IPC_POS].path,
			&ns_ctx_list[NS_NET_POS].path,
			&ns_ctx_list[NS_PID_POS].path,
			&ns_ctx_list[NS_USR_POS].path
	);
	argv += optind;
	//bb_error_msg("opts:0x%x", opts);
	//bb_error_msg("mount:%s", ns_ctx_list[NS_MNT_POS].path);
	//bb_error_msg("proc_mnt_target:%s", proc_mnt_target);
	//bb_error_msg("prop_str:%s", prop_str);
	//bb_error_msg("setgrp_str:%s", setgrp_str);
	//exit(1);

	if (setgrp_str) {
		if (strcmp(setgrp_str, "allow") == 0) {
			if (opts & OPT_map_root) {
				bb_error_msg_and_die(
					"--setgroups=allow and --map-root-user "
					"are mutually exclusive"
				);
			}
		} else {
			/* It's not "allow", must be "deny" */
			if (strcmp(setgrp_str, "deny") != 0)
				bb_error_msg_and_die("unrecognized: --%s=%s",
					"setgroups", setgrp_str);
		}
	}

	unsflags = 0;
	need_mount = 0;
	for (i = 0; i < NS_COUNT; i++) {
		const struct namespace_descr *ns = &ns_list[i];
		struct namespace_ctx *ns_ctx = &ns_ctx_list[i];

		if (opts & (1 << i))
			unsflags |= ns->flag;

		need_mount |= (uintptr_t)(ns_ctx->path);
	}
	/* need_mount != 0 if at least one FILE was given */

	prop_flags = MS_REC | MS_PRIVATE;
	/* Silently ignore --propagation if --mount is not requested. */
	if (opts & OPT_mount)
		prop_flags = parse_propagation(prop_str);

	/*
	 * Special case: if we were requested to unshare the mount namespace
	 * AND to make any namespace persistent (by bind mounting it) we need
	 * to spawn a child process which will wait for the parent to call
	 * unshare(), then mount parent's namespaces while still in the
	 * previous namespace.
	 */
	fdp.wr = -1;
	if (need_mount && (opts & OPT_mount)) {
		/*
		 * Can't use getppid() in child, as we can be unsharing the
		 * pid namespace.
		 */
		pid_t ppid = getpid();

		xpiped_pair(fdp);

		child = xfork();
		if (child == 0) {
			/* Child */
			close(fdp.wr);

			/* Wait until parent calls unshare() */
			read(fdp.rd, ns_ctx_list, 1); /* ...using bogus buffer */
			/*close(fdp.rd);*/

			/* Mount parent's unshared namespaces. */
			mount_namespaces(ppid, ns_ctx_list);
			return EXIT_SUCCESS;
		}
		/* Parent continues */
	}

	if (unshare(unsflags) != 0)
		bb_perror_msg_and_die("unshare(0x%x)", unsflags);

	if (fdp.wr >= 0) {
		close(fdp.wr); /* Release child */
		close(fdp.rd); /* should close fd, to not confuse exec'ed PROG */
	}

	if (need_mount) {
		/* Wait for the child to finish mounting the namespaces. */
		if (opts & OPT_mount) {
			int exit_status = wait_for_exitstatus(child);
			if (WIFEXITED(exit_status) &&
			    WEXITSTATUS(exit_status) != EXIT_SUCCESS)
				return WEXITSTATUS(exit_status);
		} else {
			/*
			 * Regular way - we were requested to mount some other
			 * namespaces: mount them after the call to unshare().
			 */
			mount_namespaces(getpid(), ns_ctx_list);
		}
	}

	/*
	 * When we're unsharing the pid namespace, it's not the process that
	 * calls unshare() that is put into the new namespace, but its first
	 * child. The user may want to use this option to spawn a new process
	 * that'll become PID 1 in this new namespace.
	 */
	if (opts & OPT_fork) {
		xvfork_parent_waits_and_exits();
		/* Child continues */
	}

	if (opts & OPT_map_root) {
		char uidmap_buf[sizeof("%u 0 1") + sizeof(int)*3];

		/*
		 * Since Linux 3.19 unprivileged writing of /proc/self/gid_map
		 * has been disabled unless /proc/self/setgroups is written
		 * first to permanently disable the ability to call setgroups
		 * in that user namespace.
		 */
		xopen_xwrite_close(PATH_PROC_SETGROUPS, "deny");
		sprintf(uidmap_buf, "%u 0 1", (unsigned)reuid);
		xopen_xwrite_close(PATH_PROC_UIDMAP, uidmap_buf);
		sprintf(uidmap_buf, "%u 0 1", (unsigned)regid);
		xopen_xwrite_close(PATH_PROC_GIDMAP, uidmap_buf);
	} else
	if (setgrp_str) {
		/* Write "allow" or "deny" */
		xopen_xwrite_close(PATH_PROC_SETGROUPS, setgrp_str);
	}

	if (opts & OPT_mount) {
		mount_or_die("none", "/", NULL, prop_flags);
	}

	if (opts & OPT_mount_proc) {
		/*
		 * When creating a new pid namespace, we might want the pid
		 * subdirectories in /proc to remain consistent with the new
		 * process IDs. Without --mount-proc the pids in /proc would
		 * still reflect the old pid namespace. This is why we make
		 * /proc private here and then do a fresh mount.
		 */
		mount_or_die("none", proc_mnt_target, NULL, MS_PRIVATE | MS_REC);
		mount_or_die("proc", proc_mnt_target, "proc", MS_NOSUID | MS_NOEXEC | MS_NODEV);
	}

	exec_prog_or_SHELL(argv);
}